Explainablity and transparency are key ideals-straight from Plato's realm of the forms-that developers must hold themselves to in order to facilitate proper ethical and sociological inspection of AI models. At digiLab, this is our bread and butter. This is why we have been registered onto the Ethical AI Database (EAIDB).
We look forward to contributing to this roster of companies, and hopefully, through this new collaboration, continue to find safe and ethical ways in which to deploy AI. Everyone deserves to understand and fully participate in the technologies that they are surrounded by, and lay educated claim to which trade-offs, if any, need to be made between efficiency, data privacy and security.
Less 'what do you think?', more 'what's the machine thinking?'
Ethical decision making is now part of AI algorithms, whether desired or not. Do advance machine learning models facilitate human choice, or is morality in the hands of the machines?
It's not easy to say without knowing. At digiLab we think we have an approach that helps:
- Equip deep learning models with statistical uncertainty. If the model were to change its own mind a bit, what effect would that have? The output is then a (probability) distribution of possible outcomes, like exploring the multiverse of possible realities.
- Propagate the relevance of AI decision making back to the start--to the input data it's assessing. In this was we shine a light on the impactful features in the dataset (big ears, pointy hat) that most contribute to a prediction (elf).
We believe that mathematics holds the answers to review how an AI model "thinks". And we are conscious to the implications for society.
Learn more about our debate on Ethics in AI
Our very own Michelle Fabienne Bieger has taken the reins on the ethical governance of the company. She is crucial to our mission of continuing the debate on ethics and AI. Read more about her thoughts in this excellent piece.
If you're interested in the nuts and bolts of how machine learning algorithms insight artificial thought, you might be interested in our new course 'Machine Thinking', lead by Dr Andrew Corbett.
And if you are looking for a first touch on the subject, then tuck into the series of tutorials by our Co-Founder & CEO Prof. Tim Didwell: 'From Zero to Sixty - Foundations in Machine Learning'. Lots of free content to whet your appetite.